Article
Reference
On the Neg-criterion in Hungarian
PUSKAS, Genoveva
Abstract
This paper discusses several aspects of Hungarian sentential negation. On the premises that the NEG-criterion applies at S-structure in Hungarian, I show that the NEG-criterion does represent a coherent and adequate explanatory tool to understand Hungarian negation. After establishing the motivation for a functional projection FP outside the predicational part of the sentence, as a component of CP, I show that instances of the Affect criterion, like the Focus criterion and the WH-criterion also apply at S-structure in Hungarian and account very adequately for the behaviour of non-negative quantificational elements. I propose a structure which integrates the functional projection NegP and I show that the structure I adopt has the advantage of taking into account, in addition to Hungarian adult data, acquisition data. I also show that the NEG-criterion applies at S-structure in Hungarian. The paper also discusses negative phrases. I argue that they are not negative polarity items, but intrinsically negative elements, and that the NEG-criterion applies fully at S-structure in these cases as well.
PUSKAS, Genoveva. On the Neg-criterion in Hungarian. Acta Linguistica Hungarica, 1998, vol. 45, no. 1-2, p. 167-213
Available at:
http://archive-ouverte.unige.ch/unige:96349
Disclaimer: layout of this document may differ from the published version.
1 / 1
Genoveva Puskás University of Geneva Introduction
Since Pollock (1989), much attention has been paid in the literature to the position and behaviour of negative elements (see Zanuttini 1989, Laka 1990, Progovac 1993, Acquaviva 1993, Haegeman 1995 among others). A fairly general consensus arose about the fact that negation involves a functional projection and is subject to structural constraints, although there is some controversy as to its position, as discussed in e.g. Ouhalla (1990), Acquaviva (1995), Zanuttini (to appear). In this paper, I examine the properties of Hungarian negation in the framework proposed in Haegeman (1995). Haegeman (1995) argues that negative
elements are constrained by the NEG-criterion, a well-formedness condition on the occurrence of negative elements, which is an instantiation of the more general Affect criterion. On the premises that the NEG-criterion applies at S-structure in Hungarian, I will discuss several aspects of Hungarian sentential negation which seem to challenge this assertion. I will show that although the behaviour of negative elements does not, at first sight, support the above given premise, the NEG-criterion does represent a coherent and adequate explanatory tool to understand Hungarian negation.
In the first section, I give a proposal for the structure of Hungarian non-neutral
sentences. I discuss the motivation for a functional projection FP outside the predicational part of the sentence, as a component of CP, in fact. I show that instances of the Affect criterion, like the Focus criterion and the WH-criterion also apply at S-structure in Hungarian and account very adequately for the behaviour of non-negative quantificational elements.
The second section examines recent proposals as to the position and behaviour of negative elements in Hungarian, namely in the works of Piñon (1992) and Toth (1995). I propose a structure which integrates the functional projection NegP and I show that the structure I adopt has the advantage of taking into account, in addition to Hungarian adult data, acquisition data. I also show that the NEG-criterion applies at S-structure in Hungarian.
Section 3 discusses the case of negative phrases. I argue that they are not negative polarity items, but intrinsically negative elements, and that although their behaviour seems to challenge the point made in section 2, the NEG-criterion applies fully at S-structure in these cases as well.
Section 4 discusses the various occurrences of sem. I will argue that what Brody presented as being equivalent sets of data, in which nem and sem alternate, belongs in fact to two different phenomena, both being accounted for within the framework I adopt.
Finally, section 6 gives a summary of the discussions.
1. The structure of the Hungarian sentence.
Most authors agree on the fact that Hungarian sentences can be divided into two types, the neutral order sentences and the sentences involving some focus/operator position. Neutral order sentences have been assigned various forms by linguists, but the basic assumption is that they are encompassed in IP, where the constituents ordered in an SVO pattern will be
accounted for1. Therefore, a sentence like
(1) Balázs fel fedezte az olasz filmeket.
Balázs-nom part discover-pas-3s the Italian films-acc 'Balázs discovered the Italian films.'
is assigned a structure where the subject moves to an IP-initial position, and the order particle- verb is accounted for by the adjacency of the particle and the verb in a V' node, where they are base-generated (see e.g. Brody 1990, E-Kiss 1992), or by the presence of the particle in spec TP, the verb moving into T0 (see Puskás 1996b)2.
A rather general assumption is that when the sentence does not display the word order corresponding to this SVO pattern, it is a non-neutral sentence, which contains a focus-type operator (Kenesei 1986). Again, the proposals vary as to what this entails in terms of structure:
(2) AZ OLASZ FILMEKET fedezte fel Balázs.
the Italian films-acc discover-pas-3s part Balázs-nom 'It is the Italian films that Balázs discovered.'
Whereas E-Kiss (1992) proposes that focused phrases occupy spec VP, where they are assigned /checked against the [+f] feature present on the verb (3a), Brody (1990, 1995a) and Puskás (1992), based on Choe (1989), argue that focused phrases occur in the specifier of a functional projection FP, whose head F0 hosts the verb. In this case, the verb moves to F0 where it exhibits the feature [+f] (3b):
(3)a (3)b
VP FP
3 3
spec V' spec F'
focus 3 focus 3
V XP F IP
Crucially, the difference between these two approaches lies in the fact that the projection FP entails that non-neutral sentences have a (set of) functional projection(s) above IP, and that as soon as a sentence does not exhibit a neutral word-order, the IP-external projections are activated, and the verb occurs outside IP, in a non-predicative position. The movement of the verb to an IP-external position results in particle-verb inversion, as in (2) above.
In this paper, I will continue assuming that non-neutral sentences, including negative sentences, involve a functional projection FP, which is a component of CP in the sense of Rizzi (1995) and belongs to the set of non-predicative functional projections3. The fact that FP is a component of CP, in the same sense as AgrP is a component of IP has consequences on the extraction of elements from this position as well as on selection particularities4.
It is a well know fact that focus-in-situ is ungrammatical in Hungarian:
(4) * Imádja Balázs AZ OLASZ FILMEKET. adore-pres-3s Balázs-nom the Italian films-acc
Sentence (4) above can be adequately accounted for by the Focus criterion, originally
formulated in Brody (1990). I give here a revised version of the criterion (5), based on Rizzi 's (1991) WH-criterion (6):
(5) Focus criterion
a. A +f head must be in a spec-head configuration with a +f XP b. A +f XP must be in a spec-head configuration with a +f head.
(6) WH-criterion
a. A +wh X0 must be in a spec-head configuration with a +wh operator b. A +wh operator must be in a spec-head configuration with a +wh X0
Let us now examine how the Focus criterion can account for the ungrammaticality of (4) above. The verb imádja ('adores') occurs in F0. I assume that F0 contains a feature [+f] which needs to be lexically realised, and hence forces the verb to move5. Therefore, the focused phrase az olasz filmeket ('the Italian films') violates the Focus criterion: although it carries a feature [+f] , it is not in the required spec-head configuration with the relevant head, namely
F0. Example (2) above, on the other hand, exhibits a focused phrase which has moved to the preverbal position, that is to spec FP, satisfying the Focus criterion overtly. Therefore, clause b of the Focus criterion applies at S-structure. As for clause a, I argue that it also applies at S- structure:
(7) *AZ OLASZ FILMEKET Balázs imádja.
the Italian films-acc Balázs-nom adore-pres-3s
In (7), the focused phrase az olasz filmeket occurs sentence-initially, in spec FP. However, the sentence is ruled out: as the verb imádja has not moved to F0 (it follows the subject Balázs), it does not occur in the required spec-head configuration with the focused phrase. The sentence is thus ruled out by clause a of the Focus criterion. The criterion applies fully at S-structure in Hungarian.
It is also well known that wh-phrases occur in a verb-adjacent position. Consider the following:
(8)a. Melyik filmet látta Balázs Rékával?
which film-acc see-pas-3s Balázs-nom Réka-instr 'Which film did Balázs see with Réka?'
b. RÉKÁVAL látta Balázs az olasz filmet.
Réka-instr see-pas-3s Balázs-nom the Italian film-acc 'Balázs saw the Italian film with RÉKA.'
c. *Melyik filmet RÉKÁVAL látta Balázs?
which film-acc Réka-instr see-pas-3s Balázs-nom
In (8a), the wh-phrase melyik filmet ('which film') occurs sentence-initially, immediately to the left of the verb. The subject Balázs follows the verb: therefore, the verb has moved to an IP- external position. This looks very much like focusing. Indeed, in (8b), the focused phrase Rékával ('with Réka') occurs in the preverbal position and the subject Balázs follows the verb.
(8c), which combines wh-phrase and focused phrase is ungrammatical. Therefore, it is argued in the literature (see e.g. Horváth 1981, 1986, E-Kiss 1987) that wh-phrases occur in the same position as focused phrases. In the framework adopted here, this position is spec FP. Wh- phrases are strongly constrained in Hungarian:
(9)a. *Balázs látta melyik filmet?
Balázs-nom see-pas-3s which film-acc b. *Melyik filmet Balázs látta?
In (9a), the verb occurs in its IP-internal position, and the wh-phrase melyik filmet ('which film') occupies its canonical object position. The ungrammaticality of the sentence can be accounted for by the WH-criterion (see 6 above). Indeed, the verb (or rather the inflectional head under which the verb sits) carries a feature [+wh]. In (9a), the wh-phrase does not sit in the spec position of the head which hosts the verb. Therefore, clause b of the WH-criterion is violated. In (9b), the wh-phrase has moved to the sentence-initial position, namely to spec FP.
However, the verb does not occupy the head in which it can satisfy the spec-head requirement of the WH-criterion. (9b) violates clause a of the criterion. The conclusion is again that the
WH-criterion applies fully at S-structure in Hungarian6.
In this section, I have adopted the view that Hungarian non-neutral sentences contain a functional projection F0, which belongs to the CP-type projections. Its head F0 contains a feature [+f] which signals focus and which attracts the verb. I have argued that focused phrases move to spec FP to satisfy the spec-head requirement of the Focus criterion. It was shown that the latter applies at S-structure in Hungarian. Similarly, wh-phrases move to spec FP. I have argued that they are constrained by the WH-criterion. The behaviour of wh-phrases in Hungarian shows again that the WH-criterion applies at S-structure.
2. Sentential negation
Recent proposals assign Hungarian negative sentences a structure which includes a functional projection. This projection is claimed by Piñon (1992) to be identical to the one involved in focusing. On the other hand, Toth (1995) adopts Pollock's (1989) proposal of an IP-internal NegP.
2.1 Piñon (1992)
Piñon argues that the negative marker nem is of the category 0. is the head of P, "a functional projection whose head may contain (at least) tense and negation and whose
specifier is an A'-position. This will contrast with Brody's (1990) F(ocus)P, which I believe is simply too narrow a construct." (Piñon 1992:106, fn5). Piñon also hypothesises that P is a projection for tense. He argues that it is equivalent to TP, but that it is optional. Therefore, it is the projection whose head can host the verb and whose specifier can be involved in focusing.
In short, Piñon argues that when a constituent is focused, it occurs in spec P, and the verb moves to 0, to have its tense features checked in that position (the reader is referred to Piñon (1992) for a detailed discussion). When the sentence contains a negative marker, it occurs as one of the heads of a bi-cephalic functional projection on top of VP. Piñon argues that P is in
fact an extended projection, which might have several heads with one specifier position.
Therefore, the negative sentence in (10a) will have the representation in (10b):
(10)a. MARI nem jött el tegnap.
Mary NEG came PV yesterday
'It is Mary who didn't come yesterday.'
b. P
3
spec '
MARI 3
0 P
Nem 3
0 VP
Jött rp
V0 XP XP
3 t tegnap
XP V0
el t
The specifier of P can host a focused phrase, but as P is not inherently a focus projection, the structure accommodates also sentences containing a negative marker without focused phrase, as in (11a,b) below:
(11)a. Holnap nem utazom el tomorrow NEG depart-1SG PV 'Tomorrow I'm not leaving.'
b. P 3
XP '
[e] 3
0 P
nem 3
0 VP
utazom g V 3
XP V
el t
Although Piñon's structure seems to account quite easily for the above given facts, I will argue that the analysis does not reflect the exact nature of sentence negation. As discussed in Puskás (1994), wh-movement type extraction is blocked by the presence of an intervening negation:
(12)a. Miért gondolod hogy sírt Réka?
why think-pres-2s that cry-pas-3s Réka-nom 'Why do you think that Réka cried?'
b. Miért nem gondolod hogy sírt Réka?
why neg think-pres-2s that cry-pas-3s Réka-nom 'Why don't you think that Réka cried?'
In (12a), miért ('why') can be construed with the matrix or with the embedded clause. In (12b), on the other hand, only the matrix clause reading is available. Following Rizzi (1990), this is due to the fact that an intervening position blocks the long construal. This intervening position is an A'-position, which blocks the antecedent-government relation between the wh-adjunct and its trace. Anticipating the discussion in section 2.4, I assume that this A'-position is the specifier of NegP7.
In the structure given in Piñon (1992), the negative marker nem is one of the heads of
P, and wh-phrases are moved to the specifier of P. If we adopt this structure, the wh-phrase miért ('why') in (12) above occupies spec P, and nem occurs in 0. However, the contrast between (12a) and (12b) cannot be accounted for: indeed, there is no blocking A'-position which intervenes in either of the cases. (12b) should be just as good as (12a)8. Therefore, I will conclude that NegP is an independent functional projection, lower than spec FP, and whose head carries the feature [+neg], realised in Hungarian as nem.
2.2 Toth (1995)
Toth (1995) discusses the licensing of negative polarity items in Hungarian. She argues that negative polarity items can be divided into two classes: the se-NPIs which start with the prefix se , as senki ('nobody'), semmi ('nothing'), etc., and the vala-NPI's which start with vala, as in valaki ('somebody'), valami ('something'). She observes that se-NPIs occur only with
clausemate negation. I will come back to these in section 2.4 below. As for the structure of negative sentences, Toth argues that the clause contains a NegP. Indeed, she observes that negation blocks long wh-movement and long focusing9:
(13)a. Mikor gondolod hogy megérkezik?
[when think-pres-2s that part arrive-pres-3s]
'When do you think he comes?'
b. *Mikor nem gondolod hogy megérkezik?
[when neg think-pres-3s that part arrive-pres-3s]
'When don't you think he comes?' (14)a. KÉTHETENTE szeretném, ha jönnél.
[by two weeks like-cond-1s if come-subj-2s]
'It is every two weeks that I would like you to come.' b. *KÉTHETENTE nem szeretném, ha jönnél.
[by two weeks neg like-cond-1s if come-subj-2s]
'It is every two weeks that I would not like you to come' [Toth 1995]
In (13a) and (14a), the wh-phrase mikor ('when') and the focused phrase kéthetente ('every two weeks') have been extracted from the lower clause. (13b) and (14b) differ from their a
counterparts in that the main clause contains a negative marker nem. Toth concludes that given Relativized Minimality, the antecedent government relation between the moved wh- /focused phrase and its trace is blocked as negative clauses contain a NegP whose specifier is filled at LF at the latest.
Therefore, Toth assigns the negative sentence in (15a) the structure given in (15b):
(15)a. János nem látott senkit.
[János neg see-pas-3s nobody-acc]
'John didn't see anyone.'
b. [AgrP János [Agr' nem látotti [NegP Op ti [TP [VP ti senkit ]]
Toth proposes a functional projection NegP inside IP, and she assumes that the verb (with the negative head nem) moves to Agr0. The negative phrase senkit sits in its base position, i.e.
inside VP. The Op in spec NegP is a null negative operator which is licensed by the overt negative element nem. The operator binds the negative phrase senkit ('nobody'), and assigns it sentential scope.
It is difficult to discuss Toth's proposal in detail as the set of data is not complete.
However, I will argue against the representation given in (15b) in which the sentence containing a negative element corresponds to the structure of neutral order sentences.
Consider the following pair:
(16)a. Réka nem jött el.
Réka-nom neg come-pas-3s part 'Réka did not come.'
b. *Réka nem el jött.
In (16a), the subject occurs sentence-initially. However, the sentence does not have a neutral word-order, as the particle follows the verb. Recall (see section 1) that particle-verb inversion occurs when the verb is raised to an IP-external position, that is in non-neutral sentences.
Therefore, I will conclude that negative sentences are non-neutral, and involve the IP-external functional projections.
2.3 V-movement from NegP to FP
In Puskás (1994), I argued that the structure of Hungarian negative sentences contains a projection NegP, located inside IP, between AgrP and TP. I would like to maintain the basic claim put forth in that paper, namely that negative sentences contain a NegP inside IP. I will continue assuming that NegP occurs above TP, inside IP. Indeed, I will argue, against Piñon (1992), that the functional projection responsible for negation is inside IP and that further movement to a higher position is motivated by other factors, like the [+f] feature on F0 and the Focus criterion. Like in other non-neutral sentences, the CP level is triggered and FP is
projected. As F0 contains a feature [+f] which needs to be lexicalized (see section 1), the verb moves to F0. On its way, its passes through Neg0 and takes the negative marker nem along.
The latter being a clitic, it attaches to the verb and moves along with it10. The representation of (16a) repeated here, will then be (17)11: (16)a. Réka nem jött el.
Réka-nom neg come-pas-3s part
'Réka did not come.' (17). FP
3 F'
3
F0 NegP
nem jötti/j p Neg' 3
Neg0 TP ti/j 3
el T'
3
T0 AgrSP
ti ...
The subject Réka occupies the Topic position, above FP. The verb moves through the heads of IP up to F0. On its way, it takes nem along. Thus the complex nem+verb appears under F0. I assume that in the cases where spec FP is lexically empty, the Focus criterion is satisfied by a null operator in spec FP.
The movement of nem+verb to F0 is also attested by the following examples:
(18)a. Melyik filmet nem látta Balázs?
which film-acc neg see-pas-3s Balázs-nom 'Which film did Balázs not see?'
b. *Melyik filmet Balázs nem látta?
It was shown in section 1 that wh-phrases occur in the same position as focused phrases, that is in spec FP. In (18a) above, the wh-phrase melyik filmet ('which film') occurs sentence initially, left adjacent to the negative marker. (18b) in which no verb movement takes place, violates the WH-criterion: as the WH-criterion requires that the head carrying [+wh] be in a spec-head configuration with the wh-phrase, the verbal complex nem+V which carries this feature, should occur in the head whose specifier hosts the wh-phrase. Therefore, in (18a), it is the unit formed by the negative marker nem and the verb which sits in F0.
In this section, I have shown that an analysis in which NegP (or any negative projection) appears outside IP cannot account for some of the observed facts, namely the asymmetries in extraction across a negative sentence. On the other hand, negative sentences
were shown not to be neutral order sentences. Therefore, I have proposed that the functional projection where negation is generated, that is NegP, does occur inside IP, but that the negative marker nem occurs in an IP-external position because it is a clitic which attaches onto the verb. The movement of the verb to F0, due to the constraints related to FP, forces the negative marker to appear in F0 as well, yielding the non-neutral surface order.
2.4 The NEG-criterion
Haegeman (1995) notes that negative elements trigger subject-auxiliary inversion and inner- island effects, among others. This means that in many respects, they behave like wh-operators or other affective operators (in Klima's 1964 sense). On the basis of Rizzi's WH-criterion, Haegeman and Zanuttini (1991) formulate the licensing condition on negative elements in the
NEG-criterion:
(19)a. a NEG-operator must be in a spec-head configuration with an X0 [+neg]
b. an X0 [+neg] must be in a spec-head configuration with a NEG-operator.
where the following definitions apply:
(19)c. a. NEG-operator: a negative phrase in a scope position.
b. scope position: left peripheral A'-position (an XP-adjoined position or a specifier position.
Haegeman (1995) gives the following West Flemish examples:
(20)a. da Valère dies boeken nie an zen voader getoogd (en)-oat.
that Valère those books not to his father shown en-had 'that Valère had not shown these books to his father.'
b. Valère en-eet nie s'oavends.
Valère en eats not evening's 'Valère does not eat in the evenings.'
For clarity's sake, I give in (21a,b) below a simplified bracketed representation of (20a,b):
(21)a. [CPda [AgrPValère dies boeken [NegP nie ti an zen voader getoogd [Agr (en)i-oat.]]]
b. [CPValère [C eni-eet [IP [NegP nie ti s'oavends.]]]
Haegeman (1995) argues that the negative marker en cliticizes onto the verb and moves along with it to Agr0 in (21a), and to C0in (21b). She assumes that the negative operator nie is base- generated in spec NegP. Therefore, it will enter the required spec-head relation with the trace of the negative head en, which cliticizes onto the moved verb. So the NEG-criterion can be satisfied on the trace of the negative head: although the lexical element has moved, the feature [+neg] is still available on Neg0.
Let us now examine how the NEG-criterion applies in Hungarian. Consider the data below:
(22) Nem látta Balázs ezt a filmet.
neg see-pas-3s Balázs-nom this film-acc 'Balázs didn't see this film.'
As discussed above, the functional projection which hosts sentence negation, NegP, appears inside IP. On the other hand, it was shown that nem does not occur inside IP, but in F0, cliticized onto the verb. As opposed to the West Flemish examples given above, Hungarian does not have a bi-partite bare negation: there is no overt negative operator.
I will propose that the NEG-criterion is satisfied by a null negative operator. Following Rizzi (1990a) and Haegeman (1995), I will adopt the idea that negative operators occur in contexts where negation induces inner island effects (examples from Haegeman 1995):
(23)a. Perche hai detto che Gianni e partito?
why have you said that Gianni is left 'Why did you say that Gianni has left?'
= 1.'What is the reason which made you say that Gianni left'
= 2. 'You said that Gianni left for which reason' b. Perche non hai detto che Gianni e partito?
why non have you said that Gianni is left 'Why did you not say that Gianni has left?'
='What is the reason which made you not say that Gianni has left'
In (23a), the adjunct perchè can be construed either with the higher clause or with the lower one. In (23b), on the other hand, the construal with the lower clause is lost. Haegeman argues that " in terms of a Relativized Minimality account, the intervening null operator in
[spec,NegP] blocks the antecedent-government relation between perchè and its trace in the lower clause" (Haegeman 1995:201). Note that it is not the negative marker non as such which blocks the relation, as non is a head which cliticizes onto the verb. The blocking
element is indeed the null operator present in spec NegP. Hungarian negative sentences present the same contrasts:
(24)a. Miért mondtad hogy sírt Réka?
why say-pas-2s that cry-pas-3s Réka-nom 'Why did you say that Réka cried?'
= 1. 'Which reason made you say that Réka cried'
= 2. 'You said that Réka cried for which reason' b. Miért nem mondtad hogy sírt Réka?
why neg say-pas-2s that cry-pas-3s Réka-nom 'Why didn't you say that Réka cried?'
= 'Which reason made you not say that Réka cried'
In (24a) the wh-phrase miért ('why') can be construed either with the higher or with the lower clause. However, in (24b) only the main clause reading is available12. I will conclude that the presence of a negative operator in spec NegP blocks the antecedent-government relation. In Hungarian bare sentence negation, NegP contains a null operator. I will propose that the NEG- criterion is satisfied at the level of NegP. Indeed, the example below confirms that the surface position of nem is not involved in the NEG-criterion:
(25)a. AZ ANGOL FILMEKET nem látta Balázs.
the English films-acc neg see-pas-3s Balázs-nom 'It is the English films that Balázs didn't see.'
In (25a) above, the specifier position of FP whose head hosts nem + látta ('neg+saw') is
occupied by the focused phrase az angol filmeket ('the English films'), a non-negative element.
If the NEG-criterion applied at this level, (25a) would typically violate it. Therefore, the representation for (25a) is the following:
(25)b. FP 3
spec F'
az angol filmeket 3
F0 NegP
nem látta 3 spec Neg' OP 3
Neg0 TP
t ...
+neg
As the null operator in spec NegP is present at S-structure, the conclusion is that the NEG- criterion applies at S-structure in Hungarian. Evidence for this claim will be provided by the behaviour of negative phrases, examined in section 3 below.
2.5 Negation and acquisition
The IP-internal position of NegP is, in my view, reinforced by acquisition data. Papp (1996) examines a corpus of data of Hungarian children between 1;8 and 2;9. She observes that after a first period when the children move optionally the verb (26), (27), (28), they acquire an independent functional projection FP (29):
(26)a Zoli 1;8 itt be tudu here pref can-we 'Here we can...in' b. Zoli 1;8 ide mászom be
here-to climb-I pref 'I'm climbing in here' c. Zoli 1;8 nem el-vitte
not perf-took-3sg 'He didn't take it away (27)a. Zoli 2;0 csak senki jött...Barna bácsi
only nobody came Barna uncle 'Nobody came, uncle Barna' b. Gyuri 2;3 miér(t) be-mennek?
why pref-go-they 'Why are they going in?'
c. Gyuri 2;3 mié(rt) mért ki-esett a kereke? miért ki-esett?
why why perf fell the wheel why perf fell 'Why did its wheel fall out? Why did it fall out?' (28)a. Zoli 2;2 Mikor ad oda Barna bácsinak?
when give perf Barna uncle-dat
'When (are you) giving it to Uncle Barna?' b. Gyuri 2;3 Hol dugjam be?
where put-subj-I pref 'Where shall I put it in?' c. Moni 2;4 Hova csücsülünk le?
where-to sit-we pref
'Where are we going to sit?' (29)a. Zoli 2;2 föl kell azt <venni>
pref must this <put-inf>
'This must be put on.' b. Gyuri 2;3 oda be szoktunk menni
there pref habitual-go-inf 'We usually go in there.'
The examples in (26) show that the first stage has optional verb movement: the prefix
precedes the verb in (26a), it follows the verb in (26b) and in the negative sentence (26c), the negative marker occurs sentence initially, but the prefix precedes the verb. Papp (1996) argues that functional projections might not be present at all at this stage. The examples in (27) show that typically the various instances of the Affect-criterion are violated: (27a) is a violation of the NEG-criterion: the negative marker is not present. Note that in this case, the presence of verb movement cannot be determined, as the negative element senki ('nobody') is a subject.
(27b,c) are violations of the WH-criterion: although the wh-phrase occurs sentence-initially, the verb does not move, as attested by the order particle-verb. (28) shows verb movement with wh-phrases. Papp argues that this shows that the children have implemented the functional projection which hosts the verb and satisfies the spec-head requirement of the WH-criterion.
She argues that as wh-phrases and focused phrases are in complementary distribution, this functional projection is FP. Indeed, (29) shows that with raising verbs like kell (must), szokik ('habitually (do)'), the particle occurs in the pre-verbal focus position. Her prediction is then that the acquisition of the feature [+f] (present on F0) should result in the simultaneous
acquisition of verb raising in wh, focus and negation contexts (note that Papp assumes Piñon's
structure for negative sentences). However, the data seems to contradict this. Consider examples of negative sentences of the same period:
(30)a. Gyuri 2;3 nem el-veszi not pref take-3sg 'He won't take it away.' b. Moni 2;5 én nem be-takartam
I not pref covered-it 'I didn't cover it.' c. Eva 2;7 ne nem le-ülsz ide
don't not pref sit-you here 'No, you won't sit here.'
d. Eva 2;9 nem meg-harapta a kutya a cicát.
not pref bit-3sg the dog-nom the kitten-acc 'The dog didn't bite the kitten.'
e. Eva 2;9 nem el-veszem not pref-take-I 'I won't take it away.'
At the same age, and even later, the children who produce adult wh-questions with verb- particle inversion still produce negative sentences without inversion: in the examples above, the particle systematically precedes the verb. Papp (1996) argues that in negative sentences, children project IP, to which they adjoin the negative marker. She adds that the IP-adjunction analysis is reinforced by utterances in which the child assigns scope over the whole sentence, rather than the VP:
(31)a. Zoli 2;2 nem meg-eszi csak not pref-eat-3sg just 'it won't eat it just...'
b. én nem össze-rontom csak a I not pref-ruin-I just the 'I won't ruin it just...'
Papp argues that the structures in (31) suggest that children might overgeneralize from
contrastive negation to sentential negation. However, she notes that "analysis of the children's' sentential negation in this way would of course need phonological information from the original files" (Papp 1996:12).
On the other hand, the assumption that NegP is lower in the structure makes
interesting predictions with respect to the acquisition data above as well. As the head of NegP is realised as nem, verb movement as such is not necessary: indeed, I showed in section 2.4 above that the constraint on negation formulated as the NEG-criterion is satisfied
independently of the presence of the verb itself. So by the time children have acquired the Focus- and the WH-criterion, they have also acquired the NEG-criterion. The movement of the verb to F0 in the case of negative sentences is driven by other constraints: I argued that the feature [+f] present on F0 needs to be lexically realised. In adult language, the verb (or rather the content of the functional projection T0) qualifies as the lexical element. On the other hand, example (30d) seems to indicate that for children negative sentences are not neutral sentences either: the subject occurs post-verbally. I will propose that in child sentence negation, it is the head of NegP, nem, which moves to F0. This amounts to saying that children have not
acquired the fact that nem is a clitic. They move it as a full lexical element, generating the sentences in (30).
2.6 Summary
I have argued that the structure of a negative sentence contains a NegP, whose head, Neg0, is realised as nem. This functional projection was argued to be inside IP: indeed, it was shown that argument-adjunct asymmetries observed in extractions across negation could only be accounted for under the above assumption. Besides, child language data, in which the acquisition of the various instances of the Affect-criterion were shown to appear
simultaneously also speaks in favour of a functional projection specialised in negation which is located inside IP.
On the other hand, I have discussed that negative sentences are not neutral-order sentences. I have argued that the negative marker nem surfaces outside IP, as a consequence of its being a clitic which attaches onto the verb, the latter moving to F0, an IP-external position.
However, the NEG-criterion was shown to apply at S-structure, at the level of NegP. The trace of the negative marker enters into a spec-head relation with a null negative operator. The presence of the operator seems to be confirmed precisely by the asymmetries in wh- and focus extraction across a negation: under the assumption that there is an A'-position which blocks the antecedent-government relation between a moved adjunct and its trace, it is proposed that this A'-position is precisely spec NegP, and that it is occupied by a null operator which
prevents a trace from appearing in this position. Therefore, the NEG-criterion is satisfied at this level. Evidence for this was given by examples in which the specifier of FP, whose head hosts the negative marker, contains a non-negative phrase. As no violation arises, I concluded that the NEG-criterion does not look at this position at all.
3. The negative phrases
Sentential negation in Hungarian is always expressed by the negative marker nem. In addition, negative sentences can contain various negative phrases. These elements have the property of always occurring with the negative marker:
(32)a. Balázs nem látott semmit.
Balázs-nom neg see-pas-3s nothing-acc 'Balázs didn't see anything.'
b. *Balázs látott semmit.
(33)a. Semmit nem látott Balázs.
nothing-acc neg see-pas-3s Balázs-nom 'Balázs didn't see anything.'
b. *Semmit látott Balázs
Although these negative phrases are referred to in Toth (1995) as SE-Negative Polarity Items, their properties differ from that of polarity items: they occur only in negative contexts and they are intrinsically negative13. Indeed, Toth (1995) gives the following contrasts:
(34)a. *Pál látott senkit Paul saw nobody-acc
b. *Mária nem mondta, hogy Pál látott senkit.
Mary not said that Paul saw nobody-acc 'Mary didn't say that Paul saw anybody.'
c. *Olvasott Mária semmit?
read-past Mary nothing-acc 'Did Mary read anything?' (35)a. *Pál nem mondott valamit is
Paul not said anything-acc 'Paul didn't say anything.'
b. Pál nem mondta, hogy Mária látott valakit is.
Paul not said that Mary saw anybody-acc 'Paul did not say that Mary saw anybody.'
c. Tanultál valaha is oroszul?
studied-2sg ever Russian 'Have you ever studied Russian?'
Whereas the se-type negative elements can occur only in negative contexts, as attested by the examples in (34), the vala polarity items can occur in non-negative contexts as well (35b,c).
Besides, they cannot co-occur with the negative marker nem (35a) 14. I will conclude that the se-phrases (or negative phrases) are intrinsically negative and carry a feature [+neg], in the same way as wh-phrases are endowed with a feature [+wh].
The examples above show that the negative phrases can occur either post-verbally, as in (32a) or in a preverbal position, as in (33a). West Flemish negative operators also occur in different positions in the sentence:
(36)a. [CP Z' eni-was [IP [NegP me niets ti [VP ketent ]]]].
she en was with nothing pleased 'She was not pleased with anything.'
b. [CP Me nietsk eni-was [IP ze [NegP tk ti [VP ketent ]]]].
with nothing en was she pleased 'She was not pleased with anything.'
[examples from Haegeman 1995; brackets mine.]
In the examples above, the negative phrase me niets ('with nothing') occurs either in a position inside IP, to the left of the adjective ketent ('happy') (36a), or sentence-initially, as in (36b).
Haegeman (1995) argues that in (36a), the negative operator has scrambled out of its base position to spec NegP, where it satisfies the NEG-criterion with the trace of the negative marker en. Indeed, the latter has cliticized onto the verb which occupies C015. In (36b), the negative phrase occurs in sentence-initial position: Haegeman (1995) argues that this position is spec CP. The negative phrase enters into a spec-head relation with C, where it satisfies the
NEG-criterion with the negative marker en. Therefore, Haegeman argues that in West Flemish, the NEG-criterion has the property of being able to apply at different points in the structure . We saw above that Hungarian negative phrases can also appear in two positions: either inside IP or in a sentence-initial pre-verbal position. The question is whether the two levels indeed rely on the NEG-criterion. In the following sections, I argue that it is not the case. Section 3.1 deals with post-verbal negative phrases and section 3.2 examines preverbal negatives.
3.1 NegP and the NEG-criterion Consider the following data:
(37)a. Rékával nem beszélt meg Balázs semmit.
Réka-instr neg speak-pas-3s part Balázs-nom nothing-acc 'Balázs didn't talk over (=arrange) anything with Réka.'
b. Rékával nem beszélt meg semmit Balázs.
'id.'
(38)a. Nem beszélt meg semmit a barátaival.
neg speak-pas-3s part -nom nothing-acc the friends-instr-poss 'He didn't talk over (=arrange) anything with his friends.'
b. Nem beszélt semmit meg a barátaival.
neg speak-pas-3s nothing-acc part the friends-instr-poss 'He didn't arrange anything with his friends.'
In (37a), the subject Balázs occurs post-verbally, lower than the particle, and it is followed by the negative phrase semmit. In (37b), the negative phrase precedes the subject. I assume that the subject appears in spec AgrS, in the canonical subject position. Negative phrases of the type semmit are intrinsically negative and carry a feature [+neg] (see section 3 above). Given the NEG-criterion, one expects semmit in (37) to occupy spec NegP in order to enter a spec- head relation with a negative head. However, if semmit occupies spec NegP, we have to postulate two distinct spec NegP positions. I would like to rule this option out. I will propose that in (37a), both arguments appear in their canonical positions, respectively in spec AgrSP and spec AgrOP. In (37b), on the other hand, the subject is either in VP, or maybe extraposed, yielding a slightly less natural, almost afterthought flavour to the subject16. What comes out, under this assumption, is that the negative phrase semmit does not sit in spec NegP.
The surface position of negative phrases might at a first sight challenge the
conclusions given in section 2, in which I argue that the NEG-criterion applies at S-structure.
One direction to pursue is to look at the nature of negative phrases. Recall that the NEG- criterion is expressed in terms of negative operators, where the notion of operator crucially relies on the scope position (that is a "left-peripheral A'-position) of the negative elements. In this view, Hungarian negative phrases may rely on the functional definition of operator, and when they are not in a left-peripheral A'-position, they do not count as operators. Therefore, the NEG-criterion has nothing to say about them. Then the NEG-criterion applies at LF.
Consider now the pair in (38). The contrast between (38a) and (38b) dwells in the position of the negative phrase semmit. Whereas in (38a), it occupies a position lower than the particle, that is TP, in (38b), it occurs between the verb and the particle. This position can only be spec NegP, as it occurs between the verb in F0 and the particle in spec TP. In this position, the negative phrase clearly enters into a spec-head relation with the trace of nem. Therefore, it does satisfy the NEG-criterion. It seems then that negative phrases are allowed to move out of an argument position at S-structure. However, if they do, they can only move to a position where they do not violate the NEG-criterion. However, the conclusion that the NEG-criterion applies optionally at S-structure is not the best solution, and I would like to explore another possibility.
Haegeman (1995), proposes that the NEG-criterion can dispense with the functional definition of operators as discussed above. The distinction can be captured by the notion of (overt) chain, created by movement and which accounts for negative phrases in a scope position, and that of (representational) CHAIN, created by indexation. The notion of representational CHAIN, headed by a (non-overt) scope marker, is adopted from Brody
(1995b). Brody (1995) proposes that overt movement and covert movement can be accounted for representationally in terms of chains. Chains which represent overt movement have a contentive at the head of the chain, the foot of the chain being the trace. Covert - or LF - movement involves chains whose head is an expletive, a scope marker which occurs in a relevant position, coindexed with the contentive at the foot of the chain.
I will adopt Brody's (1995) idea of expletive chain, in the spirit the discussion of Italian negation in Haegeman (1995)17. The reader should bear in mind that in a derivational framework, expletive CHAINS do not obtain through LF-movement. Rather, they are formed at S-structure, similarly to overt expletive chains.
Using this distinction, I will propose that the NEG-criterion must in fact be satisfied by an (overt) chain, that is with an element which has moved to a scope position. Similarly to the Focus- and the WH-criterion, the NEG-criterion is satisfied at S-structure. However, as opposed to wh-questions and to focused sentences, negative sentences have an overt negative marker which realises the head of the relevant functional projection; but the element which appears in the scope position is in fact non-overt. One could argue that whereas wh-phrases are
obligatory in wh-questions, negative phrases are never required as such in a negative sentence:
they do not function like the negative element pas in French or nie in West Flemish. I will propose that the NEG-criterion is always satisfied by a null negative operator. In terms of chains, we could call it a "trivial" chain, whose only element occupies an A' scope position.
In this approach, negative phrases do not function as elements which satisfy the NEG- criterion per se. Indeed, in the sense that negative phrases are not obligatory (contrary to wh- phrases in a wh-question), they do not participate in the mechanism which checks the NEG- criterion as such. In sentences with a negative phrase, as in the case of bare sentential negation (see section 2.4), the NEG-criterion itself is satisfied by a null operator in spec NegP. On the other hand, as shown by the glosses, negative phrases do contribute to the negative meaning of the whole sentence (note that following Haegeman (1995), I consider the relation between the negative head and a negative phrase not as an instance of Negative Concord, but as an
expression of sentential negation). A negative phrase which participates in sentential negation is assumed to have sentential scope. I will propose that negative phrases which have sentential scope belong to a negative CHAIN whose foot is the overt negative phrase. The latter is an A- position18. The head of the CHAIN is an empty expletive element, which occupies a scope- position at S-structure. I will assume that this position is spec NegP, an A'-position. Therefore, although the negative element as such does not occupy a scope position, the CHAIN it forms
with the (coindexed) expletive enables it to get sentential scope: it is the CHAIN as such which counts as occupying the relevant scope position, even if the contentive does not. By the
mechanism of CHAIN-formation, negative phrases which exhibit no overt A'-movement will be able to acquire sentential scope. In other words, negative phrases in Hungarian get sentential scope because they belong to a negative CHAIN, and because the head - or a member - of this
CHAIN appears in NegP, the locus of sentential negation. I will propose that the difference between (38a) and (38b) above is not a matter of level of application of the NEG-criterion. The latter always applies at S-Structure, and is satisfied by the null operator in spec NegP. Rather, in (38a), semmit appears in a CHAIN whose head, a non-overt scope marker adjoins to spec NegP. In (38b), it is the negative phrase itself which occurs in this position. Thus the (simplified) representation for (38a,b) will be the following:
(39)a [FP nemi beszélt [NegP SMk OP ti [TP meg semmitk a barátaival ]]]
b. [FP nemi beszélt [NegP semmitk OP ti [TP meg tk a barátaival ]]]
where OP is the null negative operator and SM corresponds to the expletive scope marker.
Negative phrases are [+neg]. Therefore, the CHAIN is [+neg]. The expletive which adjoins to spec NegP is thus able to undergo neg-absorption. Although it is not the expletive as such which satisfies the NEG-criterion, it does not violate it. Besides, the fact that it undergoes neg- absorption enables the negative phrase (via the CHAIN) to get sentential scope, as it forms a negative operator with the null operator in spec NegP.
The optionality of such an alternation remains to be explained. In fact, (38b) is fine only if the negative phrase is heavily stressed19. The relation between stress and the position of negative phrases is examined in more details in the section discussing Negative Concord. It will be shown that the notion of CHAIN can account for interesting contrasts in these cases.
3.2 Pre-verbal negative phrases
The negative phrases can also occur in a preverbal position, adjacent to nem+verb. Consider the following examples:
(40)a. Semmit nem látott Balázs.
nothing-acc neg see-pas-3s Balázs-nom 'Balázs didn't see ANYTHING.'
b. *Semmit BALÁZS nem látott.
nothing-acc Balázs-nom neg see-pas-3s
'BALÁZS didn't see anything.'
c. *BALÁZS semmit nem látott.
Balázs-nom nothing-acc neg see-pas-3s 'BALÁZS didn't see anything.'
d. *Semmit Balázs nem látott.
(41)a. *Ki semmit nem látott?
who-nom nothing-acc neg see-pas-3s 'Who saw nothing?'
b. *semmit ki nem látott?
c. Ki nem látott semmit?
As shown by (40a), semmit ('nothing') can occur in a preverbal position. In Puskás (1994), it is claimed that this position is spec FP and that negative phrases satisfy the NEG-criterion with the head of FP which carries the cliticized nem, and hence negative features.
Observe that (40b), where the negative phrase is not adjacent to the verb (the focused constituent Balázs intervenes), is ungrammatical. The reverse order (40c) is also
ungrammatical. Finally, in (40d), the insertion of a non-focused constituent between the negative phrase semmit and the complex nem+verb is also impossible. Given the
ungrammaticality of (40b,c), the conclusion is that preposed negative phrases land indeed in spec FP.
This seems also to be attested by the ungrammaticality of the examples (41a,b): it was shown (see section 1) that wh-phrases land in spec FP, where they satisfy the WH-criterion.
Therefore, the co-occurrence of a negative phrase and of a wh-phrase is ungrammatical, in any order: whereas (41b) is excluded by the WH-criterion (the negative phrase intervenes between the wh-phrase and the verb carrying [+wh]), (41b), where the negative phrase precedes the wh- phrase is also ruled out20.
The only possible combination of a negative- and a wh-phrase is (41c), where the wh- phrase sits in spec FP and the negative phrase in its IP-internal position. Indeed, the WH- criterion can only be satisfied in FP, where the verb carrying the feature [+wh] occurs. On the other hand, the NEG-criterion was shown (section 3.1 above) to apply in NegP, between the head Neg0 which hosts the trace of nem and the null negative operator. I will argue that in fact, it is satisfied solely at the level of the Neg P. Haegeman (1995) argues that the NEG-criterion cannot be satisfied by the trace of a negative phrase. Therefore, she concludes that in West Flemish, the NEG-criterion is satisfied in two different positions, where the actual negative phrases occur (see section 3 above). But, as discussed above, the NEG-criterion is in fact satisfied by the null operator in spec NegP. There is no need for it to be satisfied elsewhere. In fact, such a requirement would even be redundant. The position of negative phrases in
independent from the satisfaction of the NEG-criterion itself.
Therefore, I would like to argue that in the case of (41c), the movement of semmit to an IP-external position is motivated not by the NEG-criterion, but by the Focus criterion, independently of its negative component. I will assume that in this case, as it is a focused negative phrase, it carries [+f] which must satisfy the Focus criterion: when the negative phrase occurs in spec FP, it carries a strong emphatic stress. That a non-negative focused phrase and a negative focused phrase cannot co-occur in spec FP (40b,c) can be accounted for as they both have a focus feature: as there can be only one (type of) focus operator, on e of them will be excluded in the preverbal position. Indeed, there can be no absorption between a negative and a non-negative operator, hence no possibility of being interpreted as one
operator. A negative phrase moves optionally to spec FP, and if it does, it is because it requires a focus interpretation.
The sentential scope of the negative phrase was argued to obtain through the presence of a scope marker, an expletive element member of the negative CHAIN in spec NegP.
Negative phrases which occur on spec FP belong to a chain/CHAIN which enables them to get sentential scope. In the case of a movement to spec FP, the head of the chain, that is the element which occurs at the top of the chain is the negative phrase in spec FP: it occupies a scope position. But I will assume that the chain must contain a member (i.e. a trace in this case) in spec NegP in order for the phrase to attain sentential scope as a negative element: the negative chain is licensed in spec NegP. Therefore, the representation of (40a) will be the following:
(42) [FP semmiti nemj látott [NegP ti OP [Neg tj [ Balázs ti ]]
3.3 Negative concord
Consider the following examples:
(43)a. Balázs nem beszélt senkivel semmiröl.
Balázs-nom neg speak-pas-3s nobody-instr nothing-delat.
'Balázs didn't speak about anything with anybody.' b. Balázs nem beszélt semmiröl senkivel.
'id.'
(44)a. Nem beszélt Balázs senkivel semmiröl.
neg speak-pas-3s Balázs-nom nobody-instr nothing-delat.
'Balázs didn't speak about anything with anybody.' b. Nem beszélt senkivel Balázs semmiröl.
'id.'
The examples in (43), (44) above contain two negative phrases, senkivel ('with nobody') and semmiröl ('about nothing'). As shown by the glosses, these examples all exhibit Negative Concord (NC). This means that the various negative elements all contribute to a unique negative force in the sentence (see Zanuttini 1989). Haegeman and Zanuttini (1991) observe that NC applies to West Flemish under given constraints. Indeed, in order to enter NC, negative phrases must be scrambled out of the VP. If they do not, the result is a double negation (DN) reading:
(45)a. da Valère niemand nie (en-)kent that Valère nobody not en-knows 'that Valère does not know anybody.' b. da Valère nie niemand (en-)kent
'that Valère doesn't know nobody' (DN)
In (45a), niemand ('nobody') precedes nie, which sits in spec NegP. In (45b), the same negative phrase follows the nie, and the reading is that of a double negation. Haegeman and Zanuttini (1991) conclude that in (45a), the negative phrase has scrambled out of the VP and adjoins to NegP, whereas in (45b), it stays inside VP.
Haegeman (1995) argues that the scrambling of the negative phrases follows from the fact that the NEG-criterion applies at S-structure in West Flemish. Therefore, all negative phrases which occur in spec NegP enter a NC relation with the negative element nie, as in (45a). On the other hand, when negative phrases are not scrambled out of the VP, they cannot have sentential scope and fail to enter into the NC relation.
As shown by examples (43) and (44) above, Hungarian negative phrases also enter into a NC relation. However, the following contrast is interesting
:
(46)a. SENKIVEL nem beszélt SEMMIRÖL. nobody-instr neg speak-pas-3s nothing-delat.
'He didn't speak with anybody about anything.' b. SENKIVEL nem beszélt semmiröl.
'He did not speak with anybody about nothing' (DN)
In the pair above, E-Kiss (p.c) notes that the NC reading obtains when the lower negative phrase receives stress, as in (46a). Whereas in (46b), the interpretation is that of a double negation (DN). This is due to the fact that the negative phrase semmiröl is not stressed.
Although the DN reading does not seem to be available to all speakers, as pointed out by a reviewer, it seems that at least a subgroup of the native speakers does accept the contrast in
(46) above. Objections to DN readings seem to be raised by the observation that Hungarian is an NC language (see e.g. Newson 1994, Toth 1995). However, as discussed in section 3.1.
above, I take NC to apply only to the cases where several negative phrases contribute to the negative meaning of the sentence, the relation between one negative phrase and the negative marker being of a different type. Given this, the contrast in (45), to the extent that it is accepted by some speakers, is indeed a contrast with respect to NC versus DN
interpretations21. In fact, all examples containing negative phrases and which yield a NC reading require a stress, although the latter in not always prominent:
(47) Nem beszélt SENKIVEL Balázs SEMMIRÖL. neg speak-pas-3s nobody-instr Balázs-nom nothing-delat 'Balázs didn't speak about anything with anybody.'
The examples above show that the presence of a stress is a condition on the NC reading of the negative phrases. In other words, stress enables the negative phrases to reach sentential scope.
Brody (1990) also notes that "neg-phrases are always stressed in post verbal position: we have taken this throughout as indicating LF movement." (Brody 1990:224). In this approach, one will conclude that whereas negative phrases which enter into a NC move at LF, those who do not will stay in-situ. I would like to argue that it is not the case.
As discussed in section 3.1. above, negative elements which have sentential scope, and hence are able to enter into NC stay in a position which is not a scope position: as opposed to West Flemish negative phrases, they do not scramble out to spec NegP. On the other hand, they show some kind of contrast with negative phrases which do not carry stress and yield DN readings. One way out would be to say that at LF, the stressed phrases move, as opposed to the unstressed ones. However, this raises again the problem that negative phrases can move at S-structure, a fact which basically amounts to entering into a NC with the negative head. But they can also wait until LF. This is an undesirable result, as it leaves the question of the level of application of the NEG-criterion and that of scope assignment open.
I will argue that the difference between NC sentences like (45a) and DN sentences as in (45b) lies in the fact that negative phrases which result in a DN reading do not belong to a negative chain/CHAIN as discussed in section (3.1): they are in no way related to the position in NegP. Therefore, they cannot reach sentential scope: indeed, the lack of NC reading shows that their scope is not that of the negative marker and/or other negative phrases. Typically, they occupy an A-position which cannot allow them to qualify as operators. Thus, the representation of (45a) and (45b) will be, respectively, (47a) and (47b), as far as the lower negative phrase is concerned:
(47)a. [FP Senkivel [F0nemi beszélt [IP [Neg SMkOP ti [ SEMMIRÖLk ]]]]]
b. [FP Senkivel [F0nemi beszélt [IP [NegP OP ti [ semmiröl ]]]]].
In (47a), semmiröl belongs to a CHAIN <SM; semmiröl>; the same negative phrase in (47b) does not appear in a CHAIN. The fact that in (47a), the negative phrase belongs to a negative
CHAIN whose head is adjoined to spec NegP enables it, or more precisely the CHAIN, to reach sentential scope. Such a possibility is not available in (47b), where the negative phrase occupies an A-position and does not belong to a CHAIN/chain which has access to spec NegP.
Consider now the following:
(48) Nem beszélt SEMMIT meg SENKIVEL
neg speak-pas-3s nothing-acc part nobody-instr 'He didn't discuss anything with anybody.'
In (48) above, the contentive semmit ('nothing') precedes the particle. It occupies spec NegP.
As noted in section 3.1 above, it is heavily stressed. It forms a negative chain with its trace. In this case, the contentive occurs at the head of the chain <semmit; t>. On the other hand, senkivel does not appear in spec NegP. However, it contributes to the NC reading in the same way as the other negative phrase. As discussed above, it is part of a CHAIN <SM; senkivel>. I will propose that the NC reading obtains because the two chains "intersect" in spec NegP, where they get sentential scope and can undergo absorption. As discussed above, negative elements adjoined to spec NegP are able to undergo neg-absorption, forming a unique negative operator which is interpreted as a unique sentential negation reading. As negative phrases carry a feature [+neg]. I will assume that in the case of NC, the feature [+neg] is transmitted to the expletive head of the chain: thus the whole chain is [+neg]. It enables the assignment of sentential scope to the negative phrase, via the element in spec NegP.
I will conclude that in a negative chain/CHAIN, if the contentive is at the head, it is stressed. On the other hand, if it is at the foot, the stress is passed on to it: I assume that the reason is that expletives do not realise stress. Note, however, that this stress cannot be
assimilated with the stress assigned in spec FP, which has a strong emphatic or identificational reading. Therefore, it cannot be argued that these negative phrases carry [+f].22.
Consider now the following:
(49)a. SENKIVEL SEMMIRÖL nem beszélt Balázs.
nobody-instr nothing-delat neg speak-pas-3s Balázs-nom 'Balázs didn't speak with anybody about anything.'
b. Balázs SENKIVELSEMMIRÖL nem beszélt.
'id'
c. *SENKIVEL Balázs SEMMIRÖL nem beszélt.
The examples above show that more than one negative phrase can occur pre-verbally. In this case, they all enter into a NC relation. They might be preceded by a constituent (49b), which occurs in the topic position, but they cannot be separated by a constituent (49c). The negative phrases must be adjacent. So, as opposed to a focused phrase, negative phrases can be stacked in the focus slot. The reason is that as they enter NC, they undergo absorption: indeed, the negative phrases are interpreted as contributing to one and only one sentential negation23. I will assume that the first negative phrase sits in spec FP, and the other one(s) adjoin to spec FP:
(50)a. FP
3
spec F'
ru 3 spec senkivel F0 IP semmiröl
nem beszélt
...
As in the case of a single preverbal negative phrase, the negative phrases occur as elements of negative chains. I have argued that the NC reading obtains when negative chains are able to be assigned sentential scope in spec NegP. I propose that negative phrases in spec FP head jointly a negative chain, whose trace occurs in spec NegP. The full representation for (50a) will then be:
(50)b. [FPSenkivel semmiröli/k [nem beszélt [NegP ti/k OP [Neg t [Balázs ti tk ]]
3.4 Pre-verbal and post-verbal negative phrases
We have seen that negative phrases can either occur inside IP, in which case they enter into an expletive CHAIN whose head occurs in the relevant scope position, or in spec FP, in which case sentential scope is also assigned to a negative chain, whose head is the contentive negative phrase. Consider now a third possibility, which combines the two others:
(51)a. SENKIVEL nem beszélt Balázs SEMMIRÖL. nobody-instr neg speak-pas-3s Balázs-nom nothing-delat
